[Arkansan Raider]

William Sommerwerck wrote:
"Arkansan Raider" wrote in message
...
William Sommerwerck wrote:

All right... Explain the difference between digital and analog. I've
asked
the members of this group to do this several times, and none has ever
gotten it correct.

Rut roh, Raggy.
Yes. Yes, you have.
Mxsmanic, you need to pack a lunch--we could be here a while.
I'm filling the popcorn popper right now... ;^)

I am not going to get into a discussion. I simply want to see whether or not
he understands. Nothing more.



Oh man, you mean I gotta' put the popcorn away?

No, William, I was just giving you a hard time. No worries. *grin*

---Jeff

[[email protected]]

On Jan 31, 6:18*am, "Arny Krueger" wrote:
"Trevor" wrote in message

...











"Scott Dorsey" wrote in message
...
In the digital domain, practical perfection is extremely easy to achieve.
Once
everything is converted to numbers, you are no longer subject to the
tolerances and unpredictability of imperfect analog circuits. If the
conversion to digital is clean, perfection is trivial from that point on.
It's
all ones and zeroes, and whether the ones and zeroes are blurry or sharp
is
irrelevant, since perfection is the result either way. That's the whole
advantage of digital.

Right. *The problem is that conversion step. *That turns out to be really
pretty hard.
Only if you actually bothered to specify the problem conversion step being
the sonic/mechanical/electrical one, rather than the electrical A-D/D-A
one. The latter is pretty easy these days.

And that is the point. The best minds in the industry decided in the 60s and
70s that vinyl LP technology was pretty much at its zenith, and could not be
materially improved at any cost.

Any number of alternative technologies were tried, including several flavors
of FM coding, but none were adequate. *Analog media and media contact-based
playback had to go.

By the time the CD came out, its key technologies had already been proven
over several years in the form of the Laserdisc.

Which is horse****, because RCA came out with SelectaVision showing
that video could be stored on what amounted to a much higher density
vinyl pressed format. Further, the best mag tape was and remains
better than vinyl as an analog storage medium.

Digital unquestionably could be made cheaper in the long run-at the
cost of opening the Pandora's box that has resulted in the substantial
attenuation of the music industry through downloading and ripping CDs.

It came down to money.

[[email protected]]

On Jan 31, 8:05*am, "Arny Krueger" wrote:
"Les Cargill" wrote in message

...









Arny Krueger wrote:
*wrote in message
...

"Scott *wrote in message
...
In the digital domain, practical perfection is extremely easy to
achieve.
Once
everything is converted to numbers, you are no longer subject to the
tolerances and unpredictability of imperfect analog circuits. If the
conversion to digital is clean, perfection is trivial from that point
on.
It's
all ones and zeroes, and whether the ones and zeroes are blurry or
sharp
is
irrelevant, since perfection is the result either way. That's the
whole
advantage of digital.

Right. *The problem is that conversion step. *That turns out to be
really
pretty hard.

Only if you actually bothered to specify the problem conversion step
being
the sonic/mechanical/electrical one, rather than the electrical A-D/D-A
one. The latter is pretty easy these days.

And that is the point. The best minds in the industry decided in the 60s
and
70s that vinyl LP technology was pretty much at its zenith, and could not
be
materially improved at any cost.
But vinyl varied. Radically. Turntables/cartriges/stylii *varied. When
I stopped getting Stereo Review ( in 1979) , these things were all still
in some state of R&D - or at least new products were being sold. there
were dedicated retail outlets for the technology even in smallish
towns.

People are still doing technical measurements of LP playback gear, Miller
Audio Research for one. *Their tests don't show much improvement over how
things were 30 years ago.

Good vinyl playback is better understood today than when it was a
mainstream technology. lacking volume, getting it implemented can be
troublesome.

[Trevor]

"Mxsmanic" wrote in message
...
Seen from the digital side, that's true. Seen from the analog side, it's
never
true (not even in theory).

A given waveform has one optimal digital representation, which is
"perfect,"
because it cannot be improved upon in the digital realm. That
representation
will never be perfect from an analog standpoint, because you cannot
represent
an infinite number of points on a curve with a finite series of numbers.

What this boils down to is that, once you've developed an ADC that
produces
the optimal digital representation of a waveform, any further development
of
the ADC is a waste of time, because the digital representation can never
get
better.

So if you have a $50 ADC that produces this optimal digital
representation,
building a $5000 ADC accomplishes nothing.

Which in turn means that it's entirely possible that some modern ADCs are
now
"perfect" from a digital standpoint, because they produce optimal digital
representations of their input waveforms. Which means that arguing further
about conversions serves no purpose--they are as good as they will ever
be.

If you want to go further, you need a more detailed digital representation
(more samples with greater bit depth), and then you can build a fancier
ADC to
create the necessary numbers.


*Exactly*, and renders everything else you wrote irrelevent because you
place no constraints on your theoretical analog system. A higher precision,
higher data rate digital system will always cost more, just as a better
analog one will. Fortunately for the same level of performance, digital ones
now cost far less than analog.

Trevor.

[Trevor]

"William Sommerwerck" wrote in message
...
For any waveform, no matter how complex, there is a single, optimal
conversion
to a given implementation of the digital domain. Once you achieve that
optimal
conversion, no further improvement in the conversion is necessary or
possible.

And your proof that such conversion is possible, is...?

16 bit converters with DNR at the theoretical maximum (or VERY close to it),
less than 0.001% THD and IMD, less than 0.01dB frequency deviation in the
audio band (or well beyond it depending on sample rate) etc.

OK not perfect, but already closer than any human will ever need. Time to
worry about where the real problems remain, transducer performance, and
especially room acoustics!

Trevor.

[Trevor]

"William Sommerwerck" wrote in message
...
All right... Explain the difference between digital and analog. I've asked
the members of this group to do this several times, and none has ever
gotten
it correct.

Assuming they could be bothered wasting their time, how would YOU know?
:-)

Trevor.

[Mxsmanic]

William Sommerwerck writes:

All right... Explain the difference between digital and analog. I've asked
the members of this group to do this several times, and none has ever gotten
it correct.

There are several ways of looking at the difference.

An analog representation of a tree is a drawing. A digital representation of a
tree is the word "tree."

The analog representation of a tree has the advantage of being simple, and
since it is a direct physical model of a tree, it can be interpreted correctly
even by someone who has never seen a tree.

The digital representation is more complex, because it uses symbols that have
no intrinsic meaning. For correct interpretation, the person seeing the word
"tree" must know English and be able to read, and he must have already seen a
tree.

If a drawing of a tree (an analog representation) is copied, the copy will not
be identical to the original drawing. Each additional generation of copying
will produce a drawing that looks less and less like the original. Eventually,
the drawing will be uselessly unrecognizable. A similar problem occurs when
the drawing is transmitted somewhere, as by fax, for example. At the other end
of the transmission, the drawing isn't identical to the original. Eventually,
in both cases, noise and distortion destroy the original "signal" completely.

The word "tree" (a digital representation) can be copied and transmitted
indefinitely, and there will be no loss of information as long as the letters
T-R-E-E are still recognizable. It's easy to "regenerate the signal" and
maintain the information forever, so the digital representation can be copied
and transmitted with zero loss.

Digital ultimately works by setting an arbitrary threshold between what is
considered signal and what is considered noise. Anything in the channel below
the threshold is treated as noise and ignored. Anything in the channel above
the threshold is treated as signal. As long as the actual noise in the channel
is below the threshold, the digital data can be copied and transmitted forever
without loss. Thus, as long as the letters of the word "tree" are undistorted
enough to be recognizable as letters, the signal is above the threshold and
there is no loss. If the noise level rises so much that it distorts the
letters beyond recognition, there is a sudden, total loss of information.

Analog makes no distinction between signal and noise. Any noise in the channel
is treated as part of the signal, and since noise-free channels don't exist in
the real world, this means that any copying or transmission of an analog
signal introduces errors that cannot be detected, removed, or corrected. The
signal inevitably degrades. However, analog does have the advantage of being
able to use all the available bandwidth, whereas digital sacrifices some
bandwidth with its arbitrary threshold. Thus, in a low-noise channel, analog
can perform better than digital when digital has its threshold set way above
the actual noise.

A noise-free analog channel can carry an infinite amount of information at
infinite speed. A noise-free digital channel cannot, because the arbitrary
noise threshold sets an upper limit on information-carrying capacity. However,
since all channels are noisy in the real world, it turns out that analog
actually tends to perform worse than digital over the long term.

Digital transmission also requires encoding and decoding, whereas analog does
not. A drawing of a tree is recognizable even to someone who has never seen a
tree. The very analog representation itself provides much of the information
needed for interpretation. But digital data streams are nothing more than
symbols. In order to understand a digital data stream, both sender and
receiver must agree in advance on an encoding scheme that assigns a constant
meaning to the symbols used in the stream.

For example, an extraterrestrial looking at an LP might reason that the wavy
grooves on the LP are an analog of varying pressure waves in an atmosphere.
But an ET looking at a CD just sees a string of binary digits, and without
some key to the encoding of information in those digits, he has no way of
figuring out what they represent.

That's why we can understand cave drawings done thousands of years ago (analog
representations), but we can't always decipher written languages once the
speakers of those languages are dead (digital representations).

Another example of analog vs. digital is paintings vs. books. We have very few
usable images from the distant past because drawings and paintings are analog
representations that cannot be copied or transmitted without deterioration.
Over time, all images crumble into dust. But we have the written word (e.g.,
the Bible) that has survived for thousands of years because it is a digital
representation, and can be copied or transmitted without error. So we don't
have any decent pictures of ordinary ancient Romans, but we still have what
they wrote.

These are a few illustrations of the differences between digital and analog
(although they are really different sides of the same coin). In my experience,
most people have difficulty grasping these concepts, particularly those of the
digital domain.

[Dave C]

"Arny Krueger" wrote in message
...

"Mxsmanic" wrote in message
news
snip

There were good all-digital recordings from the onset. Does anybody
remember Telarc?


Agree with your whole post Arny. The 1979 Telarc recordings I have still
sound spectacular today. Telarc produced consistently excellent recordings
from the beginning of compact discs.

Dave


--- Posted via news://freenews.netfront.net/ - Complaints to ---

[Anahata]

On Tue, 31 Jan 2012 22:22:41 +0100, Mxsmanic wrote:

I suppose you could examine the results by hand and calculate whether or
not the numbers coming out are indeed the best representation of the
signal going in. But is that really worth it?

How do you measure the signal going in for comparison purposes?
(digitising it is not an option, obviously)

The permanent limitation with digital is the fact that it's a sampled
system, which means it's only perfect for a signal with strictly limited
bandwidth. To enforce that bandwidth limitation needs filters that are
hard to design (they aren't even easy to specify, it's all tradeoffs),
and much of the difference between converters has to do with the way that
filtering is done.

--
Anahata
--/-- http://www.treewind.co.uk
+44 (0)1638 720444

[Anahata]

On Tue, 31 Jan 2012 20:07:08 +0100, Mxsmanic wrote:

A given waveform has one optimal digital representation, which is
"perfect," because it cannot be improved upon in the digital realm.

Actually, it has an infinite number of equally correct digital
representations, depending on the sample timing relative to changes in
the input.

On top of that, the choice of optimal antialiasing filter is a judgement
call - there is no one perfect design as you have to trade sharpness of
cutoff againt length of impulse response.

--
Anahata
--/-- http://www.treewind.co.uk
+44 (0)1638 720444

[William Sommerwerck]

"Trevor" wrote in message
...
"William Sommerwerck" wrote in message
...

For any waveform, no matter how complex, there is a single, optimal
conversion to a given implementation of the digital domain. Once you
achieve that optimal conversion, no further improvement in the
conversion
is necessary or possible.

And your proof that such conversion is possible, is...?

16 bit converters with DNR at the theoretical maximum (or VERY close to
it),
less than 0.001% THD and IMD, less than 0.01dB frequency deviation in the
audio band (or well beyond it depending on sample rate) etc.
OK, not perfect, but already closer than any human will ever need.

How do you KNOW? How do you know that "small" numbers translate to
subjectively perfect performance?

[geoff]

Scott Dorsey wrote:
In article ,
Mxsmanic wrote:
Scott Dorsey writes:

Right. The problem is that conversion step. That turns out to be
really pretty hard.

Hard, perhaps, but not impossible. Since the accuracy of digital is
finite, there will be some possible analog system that can fully
exploit whatever accuracy it provides. Beyond that, no improvement
in the conversion is possible.

Right. But in the eighties we were having trouble just getting a 100
Hz square wave through the system without massive overshoot. Things
are better now.
--scott

So how does the same square wave go on playback from vinyl ? If you could
get it onto vinyl in the first place (as an audio signal).


geoff

[William Sommerwerck]

"Trevor" wrote in message
...
"William Sommerwerck" wrote in message
...

All right... Explain the difference between digital and analog. I've
asked
the members of this group to do this several times, and none has ever
gotten it correct.

Assuming they could be bothered wasting their time, how would YOU know?
:-)

Because I've thought it through, rather than blindly believing what's
written in books. For example... Many books say (or imply) that simply
sampling the signal converts it to digital. (Shall I name one?)

"I answered one question, and that is enough." I do not want to get into
another tsimmes over this.

[William Sommerwerck]

The answer to my question can be stated in two brief sentences. What
Mxsmanic has little to do with the "correct" answer.

I don't believe written language can be considered "digital". I've thought
about this for a long time, but have never come to a conclusion.


A noise-free analog channel can carry an infinite amount
of information at infinite speed.

Not at all true. Read a book on information theory.

[William Sommerwerck]

By the time the CD came out, its key technologies had already
been proven over several years in the form of the Laserdisc.

Which is horse****, because RCA came out with SelectaVision
showing that video could be stored on what amounted to a much
higher density vinyl pressed format.

Laserdisc (LV) predates SelectaVision (CED). The higher density of the CED
could have been used to record multi-channel sound as an FM signal. JVC had
such a system, Audio High Density (AHD), the audio-only version of Video
High Density (VHD). AHD was never commerciaized.


Further, the best mag tape was and remains better than vinyl as an
analog storage medium.

Actually, it's a much better playback medium.

[Les Cargill[_4_]]

wrote:
On Jan 31, 6:18 am, "Arny wrote:
snip

Digital unquestionably could be made cheaper in the long run-at the
cost of opening the Pandora's box that has resulted in the substantial
attenuation of the music industry through downloading and ripping CDs.



That may or may not be actually true. The measurement of that is
daunting, and I've not seen a lot of people defending that claim
empirically.

It came down to money.


It came down to money from reselling back catalog.

--
Les Cargill

[Arny Krueger[_4_]]

wrote in message
...
On Jan 31, 6:18 am, "Arny Krueger" wrote:
"Trevor" wrote in message
...


Only if you actually bothered to specify the problem conversion step
being
the sonic/mechanical/electrical one, rather than the electrical A-D/D-A
one. The latter is pretty easy these days.

And that is the point. The best minds in the industry decided in the 60s
and
70s that vinyl LP technology was pretty much at its zenith, and could not
be
materially improved at any cost.

Any number of alternative technologies were tried, including several
flavors
of FM coding, but none were adequate. Analog media and media contact-based
playback had to go.

By the time the CD came out, its key technologies had already been
proven
over several years in the form of the Laserdisc.

Which is horse****, because RCA came out with SelectaVision showing
that video could be stored on what amounted to a much higher density
vinyl pressed format.

You seem to be claiming that accepted historical facts are lies:

http://en.wikipedia.org/wiki/Capacit...lectronic_Disc

"CED players, from an early point in their life, appealed to a lower-income
market more than VHS, Betamax, and Laserdisc. The video quality (approx 3
MHz of luma bandwidth for CED[19]) was comparable to a VHS-SP or Betamax-II
video, but sub-par compared to Laserdisc (about 5 MHz of luma bandwidth)."

The meaning of the above is clear - Selectavision could not come close to
matching the data density of laserdisc, CD, and DVD.

Further, the best mag tape was and remains
better than vinyl as an analog storage medium.

Now that I agree with.

Digital unquestionably could be made cheaper in the long run-at the
cost of opening the Pandora's box that has resulted in the substantial
attenuation of the music industry through downloading and ripping CDs.

Again history seems to disagree with you. The sales of recorded media even
just music media has as a rule grown fantastically through the entire
digital revolution.


It came down to money.

It came down to what simply could not be done with analog.

[Arny Krueger[_4_]]

"Mxsmanic" wrote in message
...

William Sommerwerck writes:

But you claimed that digital conversion is or can be perfect.

Seen from the digital side, that's true. Seen from the analog side, it's
never
true (not even in theory).

A given waveform has one optimal digital representation, which is
"perfect,"
because it cannot be improved upon in the digital realm.

Not true. The quality of the digital representation can always be improved
on with more, larger samples. There are only practical limits to this.

That representation
will never be perfect from an analog standpoint, because you cannot
represent
an infinite number of points on a curve with a finite series of numbers.

The error here is the idea that an analog signal needs to be represented by
an infinite number of points. Information theory says otherwise. Every
analog signal is limited by its bandwidth and its dynamic range. The same
applies to digital, but in the digital domain these things have no
theoretical limits.


What this boils down to is that, once you've developed an ADC that
produces
the optimal digital representation of a waveform, any further development
of
the ADC is a waste of time, because the digital representation can never
get
better.

Agreed.

So if you have a $50 ADC that produces this optimal digital
representation,
building a $5000 ADC accomplishes nothing.

More to the point, if you have a $5 ADC or sometimes even a $0.50 ADC that
produces a digital signal that does not compromise the bandwidth and dynamic
range of the analog signal being digitized, then even the $50 DAC is simply
bad engienering unless it offers other kinds of advantages like being
radiation hardened for use in satellites.

Which in turn means that it's entirely possible that some modern ADCs are
now
"perfect" from a digital standpoint, because they produce optimal digital
representations of their input waveforms.

That would be true, depending on the analog signal being digitized. If the
analog signal comes from a cartridge that is playing back a LP, then it can
be pretty humble without degrading the source.

[Arny Krueger[_4_]]

"Scott Dorsey" wrote in message
...
In article ,
Mxsmanic wrote:
Scott Dorsey writes:

Right. The problem is that conversion step. That turns out to be
really
pretty hard.

Hard, perhaps, but not impossible. Since the accuracy of digital is
finite,
there will be some possible analog system that can fully exploit whatever
accuracy it provides. Beyond that, no improvement in the conversion is
possible.

Right. But in the eighties we were having trouble just getting a 100 Hz
square wave through the system without massive overshoot. Things are
better
now.

I hope you have a typo there, and mean 100 KHz.

[Arny Krueger[_4_]]

"Scott Dorsey" wrote in message
...
Mxsmanic wrote:

But... the Weiss converters sound a little different than the Prism
converters
which definitely sound different than the Grimm. So... if one of them is
indeed perfect, the question is which one.

That's the audiophile lore.

A few DBTs later... ;-)

[John Williamson[_2_]]

Arny Krueger wrote:
"Scott Dorsey" wrote in message
...
In article ,
Mxsmanic wrote:
Scott Dorsey writes:

Right. The problem is that conversion step. That turns out to be
really
pretty hard.
Hard, perhaps, but not impossible. Since the accuracy of digital is
finite,
there will be some possible analog system that can fully exploit whatever
accuracy it provides. Beyond that, no improvement in the conversion is
possible.

Right. But in the eighties we were having trouble just getting a 100 Hz
square wave through the system without massive overshoot. Things are
better
now.

I hope you have a typo there, and mean 100 KHz.


Nope, a 100Hz square wave very rarely came through an 80s system without
showing visible degradation on a scope. Either the systems were so
limited by bandwidth that it came out with visibly rounded edges or they
weren't and rang like a bell.

--
Tciao for Now!

John.

[Arny Krueger[_4_]]

"Mxsmanic" wrote in message
...
William Sommerwerck writes:

And your proof that such conversion is possible, is...?

I consider it self-evident. A digital data stream is an approximation of
the
waveform it represents. There will always be one--and only one--optimal
digital representation of that waveform. Once this representation is
obtained,
nothing more can be improved.

First, you have to define optimal.

For example, if we define an optimal conversion as one that is not audibly
different for one repetition of round-tripping the signal through the
digital domain, then things are pretty easy.

However, defining optimal by upping the number of back-to-back conversions
while retaining sonic transparency to 20 makes things quite a bit touger.

If we define optimal as reducing all errors including time-domain errors
like overshoot and undershoot to 0.01% then things can be really tough.

So, what is your idea of optimal?

[Arny Krueger[_4_]]

"William Sommerwerck" wrote in message
...
"Trevor" wrote in message
...
"William Sommerwerck" wrote in message
...

All right... Explain the difference between digital and analog. I've
asked
the members of this group to do this several times, and none has ever
gotten it correct.

Assuming they could be bothered wasting their time, how would YOU know?
:-)

Because I've thought it through, rather than blindly believing what's
written in books. For example... Many books say (or imply) that simply
sampling the signal converts it to digital. (Shall I name one?)

That book is obviously flawed because analog signals can be sampled and
remain in the analog domain.

We had CCD analog delays that did exactly that.

To finish the job, you have to quantize the samples. Reducing them to
numbers would be an example of quantizing samples.

[Arny Krueger[_4_]]

"William Sommerwerck" wrote in message
...
"Trevor" wrote in message
...
"William Sommerwerck" wrote in message
...

For any waveform, no matter how complex, there is a single, optimal
conversion to a given implementation of the digital domain. Once you
achieve that optimal conversion, no further improvement in the
conversion
is necessary or possible.

And your proof that such conversion is possible, is...?

16 bit converters with DNR at the theoretical maximum (or VERY close to
it),
less than 0.001% THD and IMD, less than 0.01dB frequency deviation in the
audio band (or well beyond it depending on sample rate) etc.
OK, not perfect, but already closer than any human will ever need.

How do you KNOW? How do you know that "small" numbers translate to
subjectively perfect performance?

DBTs.

[Arny Krueger[_4_]]

"John Williamson" wrote in message
...
Arny Krueger wrote:
"Scott Dorsey" wrote in message
...
In article ,
Mxsmanic wrote:
Scott Dorsey writes:

Right. The problem is that conversion step. That turns out to be
really
pretty hard.
Hard, perhaps, but not impossible. Since the accuracy of digital is
finite,
there will be some possible analog system that can fully exploit
whatever
accuracy it provides. Beyond that, no improvement in the conversion is
possible.

Right. But in the eighties we were having trouble just getting a 100 Hz
square wave through the system without massive overshoot. Things are
better
now.

I hope you have a typo there, and mean 100 KHz.

Nope, a 100Hz square wave very rarely came through an 80s system without
showing visible degradation on a scope.

If that happened, it was due to situations in the analog domain.

Would you accept a CDP 101 as an example of 80s digital?

Wanna see a 100 Hz square wave at the output of a CDP 101 and eat your hat?

I have a working CDP 101 and can easily burn the required test CD...

Either the systems were so limited by bandwidth that it came out with
visibly rounded edges or they weren't and rang like a bell.

I'm sure it happened, but that sort of performance was not inherent in the
digital technology of that time.

Besdies, lets not talk about about 2012 analog recorders do to 100 Hz square
waves!

[Arny Krueger[_4_]]

"Dave C" wrote in message
...

"Arny Krueger" wrote in message
...

"Mxsmanic" wrote in message
news
snip

There were good all-digital recordings from the onset. Does anybody
remember Telarc?


Agree with your whole post Arny. The 1979 Telarc recordings I have still
sound spectacular today. Telarc produced consistently excellent recordings
from the beginning of compact discs.

Thanks.

So many people around here seem to be dead-set on disproving a good general
rule with some easily explained exceptions.

[John Williamson[_2_]]

Arny Krueger wrote:
"John Williamson" wrote in message
...
Arny Krueger wrote:
I hope you have a typo there, and mean 100 KHz.

Nope, a 100Hz square wave very rarely came through an 80s system without
showing visible degradation on a scope.

If that happened, it was due to situations in the analog domain.

Undoubtedly so, I could generate a perfect 100Hz square wave on a CD
using a DAW even then. I even used to make my own test CDs for systems
once I could afford a CD burner.

Would you accept a CDP 101 as an example of 80s digital?

Wanna see a 100 Hz square wave at the output of a CDP 101 and eat your hat?

I have a working CDP 101 and can easily burn the required test CD...

I said "very rarely", not "never".

Either the systems were so limited by bandwidth that it came out with
visibly rounded edges or they weren't and rang like a bell.

I'm sure it happened, but that sort of performance was not inherent in the
digital technology of that time.

No, but it was inherent in the analogue bits before and after many of
the converters. There was also the one sample phase shift between
channels on many cheap CD players, due to them using one (at the time
relatively expensive) DAC and some cheap circuitry to switch it from one
channel to the other to look as though there were two.

Besdies, lets not talk about about 2012 analog recorders do to 100 Hz square
waves!

Chuckle About the same thing as 1980s ones did.

Tciao for Now!

John.

[Scott Dorsey]

geoff wrote:
Scott Dorsey wrote:
In article ,
Mxsmanic wrote:
Scott Dorsey writes:

Right. The problem is that conversion step. That turns out to be
really pretty hard.

Hard, perhaps, but not impossible. Since the accuracy of digital is
finite, there will be some possible analog system that can fully
exploit whatever accuracy it provides. Beyond that, no improvement
in the conversion is possible.

Right. But in the eighties we were having trouble just getting a 100
Hz square wave through the system without massive overshoot. Things
are better now.

So how does the same square wave go on playback from vinyl ? If you could
get it onto vinyl in the first place (as an audio signal).

You can make it look pretty good on the outer groove of a 33 if you're
careful about level. On the inner groove it gets pretty damn hard.

But... 100 Hz square wave through an ATR-100 looks nearly perfect! It is
just stunning!
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Arny Krueger[_4_]]

"John Williamson" wrote in message
...
Arny Krueger wrote:
"John Williamson" wrote in message
...
Arny Krueger wrote:
I hope you have a typo there, and mean 100 KHz.

Nope, a 100Hz square wave very rarely came through an 80s system without
showing visible degradation on a scope.

If that happened, it was due to situations in the analog domain.

Undoubtedly so, I could generate a perfect 100Hz square wave on a CD using
a DAW even then. I even used to make my own test CDs for systems once I
could afford a CD burner.

Would you accept a CDP 101 as an example of 80s digital?

Wanna see a 100 Hz square wave at the output of a CDP 101 and eat your
hat?

I have a working CDP 101 and can easily burn the required test CD...

I said "very rarely", not "never".

OK. ;-)


Either the systems were so limited by bandwidth that it came out with
visibly rounded edges or they weren't and rang like a bell.

I'm sure it happened, but that sort of performance was not inherent in
the digital technology of that time.

No, but it was inherent in the analogue bits before and after many of the
converters.

I can believe that. It is not unusual to find pro gear that is -3 dB @ 20
Hz, and that implies tilt and the like with a 100 Hz square wave.

There was also the one sample phase shift between channels on many cheap
CD players, due to them using one (at the time relatively expensive) DAC
and some cheap circuitry to switch it from one channel to the other to
look as though there were two.

CDP101 being a classic example. However that was audibly benign unless you
used things like electrically summed center channels. I did, so I came up
with a correction.

Besdies, lets not talk about about 2012 analog recorders do to 100 Hz
square waves!

Chuckle About the same thing as 1980s ones did.

Right and that's the point - development of analog tape had reached its
zenith or was very close to it by then.

[Mxsmanic]

anahata writes:

Actually, it has an infinite number of equally correct digital
representations, depending on the sample timing relative to changes in
the input.

The number of possible digital representations of a given waveform for a given
bit depth and sampling rate is finite. And for any depth, sampling rate, and
phase, there is only one optimal representation.

[Doug McDonald[_6_]]

On 2/1/2012 7:50 AM, Arny Krueger wrote:

Nope, a 100Hz square wave very rarely came through an 80s system without
showing visible degradation on a scope.

If that happened, it was due to situations in the analog domain.

Would you accept a CDP 101 as an example of 80s digital?

Wanna see a 100 Hz square wave at the output of a CDP 101 and eat your hat?

I have a working CDP 101 and can easily burn the required test CD...


A 100 square wave is going to ring on any CD player. Period.
A square wave has frequencies up forever. A CD player has
a sharp cutoff at 20 kHz, thus, ringing at 20 kHz. Sure ...
at 100Hz the ringing is not going to last long compared to
10 milliseconds ... but it will have the same amplitude
as it does in a 1kHz square wave.

Now in a properly designed SACD player with
a proper Bessel output filter or a 6 dB/octave filter it won't ring.


Doug McDonald

[Mxsmanic]

Arny Krueger writes:

Not true. The quality of the digital representation can always be improved
on with more, larger samples. There are only practical limits to this.

For any given depth, sampling rate, and phase, there is only one optimal
digital representation of a given waveform.

The error here is the idea that an analog signal needs to be represented by
an infinite number of points.

Some analog signals can be represented by a finite number of samples, but not
all. White noise cannot be represented in this way, because every point in the
waveform is random.

[Mxsmanic]

anahata writes:

How do you measure the signal going in for comparison purposes?

You could plot it on a scope, for example, or plot it on paper. That is
incidental to my point, however.

[William Sommerwerck]

"Mxsmanic" wrote in message
...
William Sommerwerck writes:

I don't believe written language can be considered "digital".

This isn't conjecture, it's fact. Written language is a classic example of
digital representation. Digital representation uses symbols that follow an
encoding scheme. Analog representation uses a physical model. Alphabets
(and other similar systems of writing) are not physical models, they are
encoded symbols.

Not at all true. Read a book on information theory.

I've read a lot of books on information theory, but this is pretty
fundamental.

Have you ever wondered how modems send 56 kbps over a telephone line that
supposedly has a bandwidth of 4800 bps?

Yes, I know how they do it.

[Mxsmanic]

geoff writes:

So how does the same square wave go on playback from vinyl? If you could
get it onto vinyl in the first place (as an audio signal).

I don't see how any true square wave could be recorded or tracked on an LP,
since it would require radial groves and instantaneous movement of the stylus.

[Mxsmanic]

William Sommerwerck writes:

I don't believe written language can be considered "digital".

This isn't conjecture, it's fact. Written language is a classic example of
digital representation. Digital representation uses symbols that follow an
encoding scheme. Analog representation uses a physical model. Alphabets (and
other similar systems of writing) are not physical models, they are encoded
symbols.

Not at all true. Read a book on information theory.

I've read a lot of books on information theory, but this is pretty
fundamental.

Have you ever wondered how modems send 56 kbps over a telephone line that
supposedly has a bandwidth of 4800 bps? Or what the difference between baud
rate and bits per second really is?

[Mxsmanic]

Arny Krueger writes:

First, you have to define optimal.

The representation that allows the closest approach to perfect reproduction of
the original signal.

So, what is your idea of optimal?

See above.

[Scott Dorsey]

John Williamson wrote:
Arny Krueger wrote:
"John Williamson" wrote in message
...
Arny Krueger wrote:
I hope you have a typo there, and mean 100 KHz.

Nope, a 100Hz square wave very rarely came through an 80s system without
showing visible degradation on a scope.

If that happened, it was due to situations in the analog domain.

Undoubtedly so, I could generate a perfect 100Hz square wave on a CD
using a DAW even then. I even used to make my own test CDs for systems
once I could afford a CD burner.

There really weren't any DAWs then, aside from the Sony thing that nobody
could afford. There was the PCM-1610 and the PCM-1630, and I defy you to
make anything even remotely like an accurate 100 Hz square wave on them.

A decade later at the beginning of the DAT era, there were some fairly
decent professional converters from folks like Apogee (who started out making
retrofit kits for the PCM-1610 and PCM-F1), but the internal converters in
most equipment were just ghastly. Except for the SV-3700 which was worse
than ghastly.

But by 1990 or so there were the beginnings of some workable DAW systems
with the AudioMedia card and the first generation Waves and Sonic gear.
And the Sonic software could actually be bit-for-bit accurate if necessary.
Also at around that time, folks started realizing that dither was important.

Would you accept a CDP 101 as an example of 80s digital?

Wanna see a 100 Hz square wave at the output of a CDP 101 and eat your hat?

I have a working CDP 101 and can easily burn the required test CD...

I said "very rarely", not "never".

I'd be curious to see what a 100 Hz square wave looked like on a CDP 101.
I never got a chance to test any of that playback gear with material that
had not already been pretty badly mutilated by poor A/Ds.

Besdies, lets not talk about about 2012 analog recorders do to 100 Hz square
waves!

Chuckle About the same thing as 1980s ones did.

There's a big difference in that regard between a 350 and an ATR-100 though.
Absolutely boggling, too.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

[Frank Stearns]

"William Sommerwerck" writes:

The answer to my question can be stated in two brief sentences. What
Mxsmanic has little to do with the "correct" answer.

I don't believe written language can be considered "digital". I've thought
about this for a long time, but have never come to a conclusion.

In the abstract, could it not?

The "analog" of the written symbols can vary quite a bit (look at my lousy
handwriting compared to someone with good penmenship; look at all the type faces one
can select in various applications; you can even WRITE IN ALL CAPS and the
symbols can still be recognized as a variant of their lower-case siblings).

But the message conveyed by all those varying symbols, say, "Mary had a Little Lamb"
is preserved. (We must be careful here. It's not specifically the message, but
rather the meaning we get from the message -- what we visualize in our heads -- when
we read "Mary had a little lamb.")

With a just a little care you can mess with what those encoding symbols look like,
copy them over and over and over, and if nothing has gone too far outside of agreed
limits you can still extract the **meaning** of "Mary had a little Lamb" from that
string of symbols.

This seems to be the frame of reference Mxsmanic was writing from. A digital signal,
which can show some wide variances in amplitude, rise and times, noise and other
distortions, can still have the content extracted if the variants don't pull it out
of a pre-agreed acceptable range.

As an abstract exercise, his assertion here of calling that "symbol encoding"
digital seems correct.

If you want to attritute some analog characteristics to a digital signal in the real
world, you're also correct, but it's what you do with that signal that counts in the
end.

As long as it doesn't deviate beyond usability, we don't care what a mangled digital
signal looks like, whereas with analog we surely would!

Perhaps my reply here isn't much of a contribution. Can you add some specific
notions as to why you would not consider written language "digital" in this
context?

Frank
Mobile Audio
--

[Scott Dorsey]

Mxsmanic wrote:
geoff writes:

So how does the same square wave go on playback from vinyl? If you could
get it onto vinyl in the first place (as an audio signal).

I don't see how any true square wave could be recorded or tracked on an LP,
since it would require radial groves and instantaneous movement of the stylus.

Actually, it doesn't, because the _velocity_ of the stylus is proportional
to the voltage, not the _position_. That's what all that RIAA equalization
stuff is about. Not that it isn't difficult because the cutting stylus and
carriage have mass, but it's possible to do a surprisingly good job in spite
of the system being severely slew-limited.

Check out the Shure S16 test record for 1 KHz square waves, which are even
harder to track properly. Mind you, that was cut at half-speed on a dmm
lathe.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."